nvdec syncpt incorporation

laying the groundwork for async gpu, although this does not fully implement async nvdec operations

src/video_core/cdma_pusher.cpp

@@ -33,8 +33,7 @@ CDmaPusher::CDmaPusher(GPU& gpu_)
     : gpu{gpu_}, nvdec_processor(std::make_shared<Nvdec>(gpu)),
       vic_processor(std::make_unique<Vic>(gpu, nvdec_processor)),
       host1x_processor(std::make_unique<Host1x>(gpu)),
-      nvdec_sync(std::make_unique<SyncptIncrManager>(gpu)),
-      vic_sync(std::make_unique<SyncptIncrManager>(gpu)) {}
+      sync_manager(std::make_unique<SyncptIncrManager>(gpu)) {}
 
 CDmaPusher::~CDmaPusher() = default;
 
@@ -110,10 +109,10 @@ void CDmaPusher::ExecuteCommand(u32 state_offset, u32 data) {
             const auto syncpoint_id = static_cast<u32>(data & 0xFF);
             const auto cond = static_cast<u32>((data >> 8) & 0xFF);
             if (cond == 0) {
-                nvdec_sync->Increment(syncpoint_id);
+                sync_manager->Increment(syncpoint_id);
             } else {
-                nvdec_sync->IncrementWhenDone(static_cast<u32>(current_class), syncpoint_id);
-                nvdec_sync->SignalDone(syncpoint_id);
+                sync_manager->SignalDone(
+                    sync_manager->IncrementWhenDone(static_cast<u32>(current_class), syncpoint_id));
             }
             break;
         }
@@ -135,10 +134,10 @@ void CDmaPusher::ExecuteCommand(u32 state_offset, u32 data) {
             const auto syncpoint_id = static_cast<u32>(data & 0xFF);
             const auto cond = static_cast<u32>((data >> 8) & 0xFF);
             if (cond == 0) {
-                vic_sync->Increment(syncpoint_id);
+                sync_manager->Increment(syncpoint_id);
             } else {
-                vic_sync->IncrementWhenDone(static_cast<u32>(current_class), syncpoint_id);
-                vic_sync->SignalDone(syncpoint_id);
+                sync_manager->SignalDone(
+                    sync_manager->IncrementWhenDone(static_cast<u32>(current_class), syncpoint_id));
             }
             break;
         }

src/video_core/cdma_pusher.h

@@ -116,12 +116,10 @@ private:
     void ThiStateWrite(ThiRegisters& state, u32 state_offset, const std::vector<u32>& arguments);
 
     GPU& gpu;
-
-    std::shared_ptr<Nvdec> nvdec_processor;
-    std::unique_ptr<Vic> vic_processor;
-    std::unique_ptr<Host1x> host1x_processor;
-    std::unique_ptr<SyncptIncrManager> nvdec_sync;
-    std::unique_ptr<SyncptIncrManager> vic_sync;
+    std::shared_ptr<Tegra::Nvdec> nvdec_processor;
+    std::unique_ptr<Tegra::Vic> vic_processor;
+    std::unique_ptr<Tegra::Host1x> host1x_processor;
+    std::unique_ptr<SyncptIncrManager> sync_manager;
     ChClassId current_class{};
     ThiRegisters vic_thi_state{};
     ThiRegisters nvdec_thi_state{};

src/video_core/command_classes/host1x.cpp

@@ -34,6 +34,8 @@ void Tegra::Host1x::ProcessMethod(Method method, const std::vector<u32>& argumen
 }
 
 void Tegra::Host1x::Execute(u32 data) {
-    // This method waits on a valid syncpoint.
-    // TODO: Implement when proper Async is in place
+    u32 syncpointId = (data & 0xFF);
+    u32 threshold = state.load_syncpoint_payload32;
+
+    gpu.WaitFence(syncpointId, threshold);
 }

src/video_core/command_classes/sync_manager.cpp

@@ -38,7 +38,7 @@ u32 SyncptIncrManager::IncrementWhenDone(u32 class_id, u32 id) {
 }
 
 void SyncptIncrManager::SignalDone(u32 handle) {
-    const auto done_incr =
+    const auto& done_incr =
         std::find_if(increments.begin(), increments.end(),
                      [handle](const SyncptIncr& incr) { return incr.id == handle; });
     if (done_incr != increments.cend()) {